Short answer: Tana's desktop app has no built-in dictation, but any system-wide voice tool types straight into its nodes. Hold a hotkey with Voice Keyboard Pro, speak your note, and release; the text lands at your cursor. Speak the content, then add tags, fields, and outline structure with the keyboard.
Tana calls itself an "everything OS," and the pitch is real: notes, tasks, people, meetings, and projects live in one outline where any node can become structured data with a supertag. The cost of that power is input. A Tana workspace rewards you for capturing more (more meeting notes, more observations, more context on every person and project), and every one of those captures is typed by hand, one 40-word-per-minute keystroke at a time.
You speak three to four times faster than you type. In a tool whose entire philosophy is "capture now, structure later," voice is the missing capture method. This guide covers how to dictate into Tana on a Mac, which parts of Tana respond well to voice and which genuinely belong to the keyboard, and how to build a capture habit where the bottleneck is your thinking rather than your fingers.
What Tana Offers Out of the Box
On mobile, Tana has a capture app that can take voice memos, and it can process recordings into text for you. It is designed for the "idea in the parking lot" scenario, and it is good at it. But it is a capture inbox, not a way of writing in your workspace: the recording gets processed and lands as a note to be filed, on the app's schedule rather than in real time at your cursor.
On the desktop, where most serious Tana work happens, there is no built-in dictation at all. When you are in the outline itself (expanding a meeting node, filling in fields on a person, writing under today's daily page) you type, or you bring your own voice tool. The good news: because Tana runs as a standard desktop app with ordinary editable text, system-wide dictation tools work in every node and field.
Two Ways to Dictate on the Desktop
macOS built-in dictation
Free and already on your Mac: enable it in System Settings → Keyboard → Dictation, click into a Tana node, press the mic key, and speak. It works for a sentence or two. Its weaknesses matter more in Tana than in most apps, though. It stops listening when you pause to think, it needs every comma and period spoken aloud, and it reliably mangles the proper nouns (people, projects, company names) that a knowledge graph is made of. A workspace full of misspelled names is a workspace whose links and references quietly break.
A hold-to-talk tool: Voice Keyboard Pro
Voice Keyboard Pro is a menu bar app built for exactly this input model. Hold a hotkey, speak for as long as you like (pauses and all), release, and the transcribed text appears at your cursor about a second later, punctuated and capitalized by its advanced AI transcription. It is system-wide, so nothing about Tana needs configuring; if the cursor is in a node, the words go in the node.
Setup: download the app, grant microphone and accessibility permissions, choose a hotkey you can comfortably hold with one hand. That last detail matters in Tana specifically, because your other hand stays on the keyboard for the structural keys, which brings us to the workflow that actually works.
The Core Workflow: Speak Content, Type Structure
Tana text comes in two kinds. There is content: the sentences that carry meaning, the meeting summary, the idea, the context on a project. And there is structure: the Tab that indents a node, the Enter that creates the next one, the # that applies a supertag, the @ that links a date or a person. The division of labor is clean:
- Voice for content. Anything that is a sentence or a paragraph, speak it. This is where the 3x speed advantage lives.
- Keyboard for structure. Tags, links, field navigation, indenting, and node creation are single keystrokes. They are already fast, and they are syntax, not prose.
In practice a captured thought looks like this: press Enter for a new node, hold the hotkey and speak the thought, release, then type #idea or #task to structure it. Two seconds of keyboard, zero seconds of typing prose. The same principle (speak the words, type the syntax) is the one we recommend for every outliner; if you have read our guides to dictating in Roam Research or dictating in Logseq, the muscle memory transfers directly. One habit to develop: dictate one node at a time rather than a monologue into a single node. Tana's power comes from granular nodes you can tag and reference individually, and a five-sentence block defeats that. Speak a thought, release, Enter, speak the next.
Where Voice Shines in Tana
The daily page
Tana opens to today, and today is where raw capture belongs. A morning planning session dictated node by node takes two or three minutes out loud. Journaling in particular changes character by voice: spoken entries come out longer, looser, and more honest than typed ones, because talking does not invite the same self-editing that a keyboard does.
Meeting notes
This is the highest-leverage use. During a call, keep your hands free: hold the hotkey, summarize the point just made in one sentence, release. You capture in near real time without the clatter of typing over the conversation. If the meeting is on your Mac, Voice Keyboard Pro's Meeting Mode goes further: it detects the meeting (including from your calendar), listens with speaker detection, and produces AI notes you can paste under the meeting node with a #meeting supertag, ready for Tana's fields and references. For the pattern of pairing live capture with structured summaries, our guide to dictating meeting minutes goes deeper.
Supertag fields
Fields are just text targets. Click into a Description, Notes, or Next steps field on any tagged node and dictate the value. Fields whose values are references (a person, a date, a status) are keyboard territory; fields whose values are sentences are voice territory. You will find the split becomes automatic within a day.
Fleshing out stubs
Every Tana workspace accumulates skeleton nodes: a person with an empty bio, a project with no context, a book note containing only the title. These stay empty because typing three sentences of context per stub is tedious. Speaking three sentences is not. A weekly ten-minute pass, dictating context into your barest nodes, compounds into the difference between a graph and a list of names.
Teaching It Your Graph's Vocabulary
A knowledge graph is proper nouns all the way down, and proper nouns are what generic dictation gets wrong. Voice Keyboard Pro's Smart Vocabulary is a personal dictionary with replacement rules: add your collaborators' names, project codenames, client companies, and field-specific terms once, and the transcription engine renders them correctly, capitalization included, from then on. This matters doubly in Tana because names are load-bearing; a reference to a misspelled person node is a reference to nobody. Replacement rules also handle shorthands: teach it that a spoken phrase should come out as your standard weekly-review template, or that "q3 roadmap" always renders in your house style.
Capture on the Phone, Structure on the Mac
The classic PKM failure is the idea that dies between where you are and where your workspace is. Tana's mobile capture app covers the recorded-memo route. Voice Keyboard Pro's iPhone keyboard covers the real-time route: it is a custom keyboard with a built-in mic button, so you can dictate directly into Tana's mobile app (or any app) and watch the text appear as you speak, rather than waiting for a recording to be processed. Speak the raw thought into your phone in the moment, then do the tagging and filing later at your Mac, where structure work belongs. Capture and organization are different jobs, and Tana's own design agrees.
Troubleshooting
- Nothing appears when you speak. The cursor must be inside a node, not merely on the Tana window. Click into the outline so you see a blinking text cursor, then dictate. For Voice Keyboard Pro, confirm accessibility permission in System Settings → Privacy & Security.
- Text lands in the wrong node. Dictated text follows the cursor. Get in the habit of creating the node first (
Enter), then speaking, rather than speaking mid-navigation. - Spoken "hashtag" does not create a supertag. Correct, and by design: dictation types words, and Tana's tag picker responds to the typed
#character with its own autocomplete. Type tags; speak sentences. - Names come out wrong. Add them to Smart Vocabulary. Five minutes with your twenty most-referenced people and projects fixes the overwhelming majority of errors.
A One-Week Adoption Plan
Voice input in a structured tool is a habit change, and habit changes fail when you try to convert everything at once. A schedule that works:
- Days 1-2: daily page only. Dictate your morning plan and end-of-day recap as individual nodes. Low stakes, high repetition, and it builds the hold-speak-release rhythm without touching your important structures.
- Day 3: load Smart Vocabulary. Skim your workspace for the names you reference most (people, projects, clients, tools) and add the top twenty. Do this before judging accuracy; most "dictation is inaccurate" verdicts are really "dictation does not know my nouns yet."
- Days 4-5: meeting notes. Bring the hotkey into two real meetings. Capture one spoken sentence per discussion point, then tag and file afterward.
- Days 6-7: backfill. Do the stub-fleshing pass described above on ten empty nodes. This is where voice stops being a capture trick and starts changing what your graph contains.
By the end of the week you will know your split. Most people settle around dictating the majority of new prose while keeping all structural work on the keyboard, which is exactly the ratio Tana's design wants.
Frequently Asked Questions
Does dictation work in Tana in the browser?
Yes. A system-wide tool types at the cursor regardless of whether Tana is running in its desktop app or a browser tab. The desktop app is the nicer home for a voice workflow simply because it keeps Tana one ⌘Tab away, but nothing about dictation itself changes.
Can I dictate Tana commands or AI prompts?
Prompts, yes: anywhere Tana expects natural-language input, spoken sentences are ideal, and they tend to be better prompts because people naturally speak with more context than they type. Commands, no: command menus respond to typed characters and keyboard navigation, and voice adds a layer of indirection to something that is already two keystrokes.
What about dictating in other languages?
Voice Keyboard Pro's transcription engine supports dozens of languages, so a bilingual workspace (say, client notes in German, personal notes in English) works without switching any settings between captures.
Is this different from using Tana's voice memos?
Different job. A recorded memo is asynchronous: you talk, a note arrives later, you file it. Hold-to-talk dictation is synchronous writing: the words appear at your cursor as part of the editing session you are already in, node by node, exactly where you put them. Most Tana users end up wanting both, with memos for the parking lot and dictation for the desk.
What It Costs, and a Note on Privacy
macOS dictation is free. Voice Keyboard Pro has a free tier with daily limits (enough to run a real week of Tana capture and see if the habit sticks), and Pro is $4.99 a month or $34.99 a year. On privacy, since a workspace like Tana holds your working mind: our server stores only operational pings, with no audio and no transcript content. What you dictate goes to your cursor and stays yours.
The Bottom Line
Tana already solved the structure problem: supertags and fields turn loose notes into a queryable graph with almost no friction. The remaining friction is getting words into the outline at all, and that is a solved problem too; it is just solved by your voice rather than your fingers. Keep the keyboard for what it is genuinely best at (tags, links, indentation, references) and hand the sentences to a hotkey. Try Voice Keyboard Pro free, open today's daily page, and dictate your first ten nodes. The graph grows faster when you stop typing it.